Similarity-based Methods Research Articles

Maximizing RAG efficiency: A comparative analysis of RAG methods

Abstract This paper addresses the optimization of retrieval-augmented generation (RAG) processes by exploring various methodologies, including advanced RAG methods. The research, driven by the need to enhance RAG processes as highlighted by recent studies, involved a grid-search optimization of 23,625 iterations. We evaluated multiple RAG methods across different vectorstores, embedding models, and large language models, using cross-domain datasets and contextual compression filters. The findings emphasize the importance of balancing context quality with similarity-based ranking methods, as well as understanding tradeoffs between similarity scores, token usage, runtime, and hardware utilization. Additionally, contextual compression filters were found to be crucial for efficient hardware utilization and reduced token consumption, despite the evident impacts on similarity scores, which may be acceptable depending on specific use cases and RAG methods.

Improved similarity-based link prediction methods in trade networks by integrating community-based popularity

Link prediction is a vital aspect of analyzing network evolution and identifying potential connections in complex networks. Previous studies have primarily focused on the information of common neighbors between nodes, often overlooking the inherent attributes of nodes. This study proposes community-based popularity, an attribute of nodes that considers changes in the neighborhood over time in conjunction with the community structure. Based on this attribute, we improve similarity-based link prediction methods. The experiments utilized unweighted directed networks from three distinct types of trade to evaluate the effectiveness of the improved link prediction methods. The training and probe sets were divided in chronological order. The experimental results show that the improved methods provide better link prediction results than the compared methods.

Mitigating Grand Challenges in Life Cycle Inventory Modeling through the Applications of Large Language Models.

The accuracy of life cycle assessment (LCA) studies is often questioned due to the two grand challenges of life cycle inventory (LCI) modeling: (1) missing foreground flow data and (2) inconsistency in background data matching. Traditional mechanistic methods (e.g., process simulation) and existing machine learning (ML) methods (e.g., similarity-based selection methods) are inadequate due to their limitations in scalability and generalizability. The large language models (LLMs) are well-positioned to address these challenges, given the massive and diverse knowledge learned through the pretraining step. Incorporating LLMs into LCI modeling can lead to the automation of inventory data curation from diverse data sources and to the implementation of a multimodal analytical capacity. In this article, we delineated the mechanisms and advantages of LLMs to addressing these two grand challenges. We also discussed the future research to enhance the use of LLMs for LCI modeling, which includes the key areas such as improving retrieval augmented generation (RAG), integration with knowledge graphs, developing prompt engineering strategies, and fine-tuning pretrained LLMs for LCI-specific tasks. The findings from our study serve as a foundation for future research on scalable and automated LCI modeling methods that can provide more appropriate data for LCA calculations.

An improved similarity-based RUL prediction method considering degradation degree of multiple condition monitoring parameters for aero-engines

Abstract The aero-engine is the heart of an airplane. Predicting the remaining useful lifetime (RUL) of an aero-engine bears great significance, not only for improving the reliability and safety of the aero-engine but also for ensuring aircraft safety and performance. However, both issues, namely the selection of uncorrelated parameters for RUL estimation and the lack of a standard theoretical methodology for Health Index (HI) construction, inevitably impact the prediction accuracy. Here, we proposed an improved similarity-based RUL prediction method considering the degradation degree of multiple condition monitoring parameters for aero-engines. This method includes the improved minimum-redundancy maximum-relevancy approach for the quantitative selection of key parameters, and the similarity matching approach which takes into account the degradation degree of multiple parameters instead of constructing HI. The effectiveness of the proposed method was evaluated on turbine engine datasets. Experimental results show that (1) compared with other feature selection methods, the root mean square error (RMSE) of the proposed method is reduced by 12.3%; (2) compared with other RUL prediction methods, the RMSE of the proposed method is smaller than most methods but without a complicated training process. The results demonstrate that the proposed method achieves highly competitive prediction performance. By employing the proposed method, it is possible to significantly reduce the risk of engine failures, thereby improving safety and economic efficiency.

A Comparative Study of Data-Driven Prognostic Approaches under Training Data Deficiency

In industrial system health management, prognostics play a crucial role in ensuring safety and enhancing system availability. While the data-driven approach is the most common for this purpose, they often face challenges due to insufficient training data. This study delves into the prognostic capabilities of four methods under the conditions of limited training datasets. The methods evaluated include two neural network-based approaches, Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks, and two similarity-based methods, Trajectory Similarity-Based Prediction (TSBP) and Data Augmentation Prognostics (DAPROG), with the last being a novel contribution from the authors. The performance of these algorithms is compared using the Commercial Modular Aero-Propulsion System Simulation (CMAPSS) datasets, which are made by simulation of turbofan engine performance degradation. To simulate real-world scenarios of data deficiency, a small fraction of the training datasets from the original dataset is chosen at random for the training, and a comprehensive assessment is conducted for each method in terms of remaining useful life prediction. The results of our study indicate that, while the Convolutional Neural Network (CNN) model generally outperforms others in terms of overall accuracy, Data Augmentation Prognostics (DAPROG) shows comparable performance in the small training dataset, being particularly effective within the range of 10% to 30%. Data Augmentation Prognostics (DAPROG) also exhibits lower variance in its predictions, suggesting a more consistent performance. This is worth highlighting, given the typical challenges associated with artificial neural network methods, such as inherent randomness, non-intuitive decision-making processes, and the complexities involved in developing optimal models.

Evolutionary History of the DD41D Family of Tc1/Mariner Transposons in Two Mayetiola Species.

Tc1/mariner elements are ubiquitous in eukaryotic genomes including insects. They are diverse and divided into families and sub-families. The DD34D family including mauritiana and irritans subfamilies have already been identified in two closely related species of Cecidomyiids M. destructor and M. hordei. In the current study the de novo and similarity-based methods allowed the identification for the first time of seven consensuses in M. destructor and two consensuses in M. hordei belonging to DD41D family whereas the in vitro method allowed the amplification of two and three elements in these two species respectively. Most of identified elements accumulated different mutations and long deletions spanning the N-terminal region of the transposase. Phylogenetic analyses showed that the DD41D elements were clustered in two groups belonging to rosa and Long-TIR subfamilies. The age estimation of the last transposition events of the identified Tc1/mariner elements in M. destructor showed different evolutionary histories. Indeed, irritans elements have oscillated between periods of silencing and reappearance while rosa and mauritiana elements have shown regular activity with large recent bursts. The study of insertion sites showed that they are mostly intronic and that some recently transposed elements occurred in genes linked to putative DNA-binding domains and enzymes involved in metabolic chains. Thus, this study gave evidence of the existence of DD41D family in two Mayetiola species and an insight on their evolutionary history.

DrugRepoBank: a comprehensive database and discovery platform for accelerating drug repositioning.

In recent years, drug repositioning has emerged as a promising alternative to the time-consuming, expensive and risky process of developing new drugs for diseases. However, the current database for drug repositioning faces several issues, including insufficient data volume, restricted data types, algorithm inaccuracies resulting from the neglect of multidimensional or heterogeneous data, a lack of systematic organization of literature data associated with drug repositioning, limited analytical capabilities and user-unfriendly webpage interfaces. Hence, we have established the first all-encompassing database called DrugRepoBank, consisting of two main modules: the 'Literature' module and the 'Prediction' module. The 'Literature' module serves as the largest repository of literature-supported drug repositioning data with experimental evidence, encompassing 169 repositioned drugs from 134 articles from 1 January 2000 to 1 July 2023. The 'Prediction' module employs 18 efficient algorithms, including similarity-based, artificial-intelligence-based, signature-based and network-based methods to predict repositioned drug candidates. The DrugRepoBank features an interactive and user-friendly web interface and offers comprehensive functionalities such as bioinformatics analysis of disease signatures. When users provide information about a drug, target or disease of interest, DrugRepoBank offers new indications and targets for the drug, proposes new drugs that bind to the target or suggests potential drugs for the queried disease. Additionally, it provides basic information about drugs, targets or diseases, along with supporting literature. We utilize three case studies to demonstrate the feasibility and effectiveness of predictively repositioned drugs within DrugRepoBank. The establishment of the DrugRepoBank database will significantly accelerate the pace of drug repositioning. Database URL: https://awi.cuhk.edu.cn/DrugRepoBank.

Prediction of Human Liver Microsome Clearance with Chirality-Focused Graph Neural Networks.

In drug candidate design, clearance is one of the most crucial pharmacokinetic parameters to consider. Recent advancements in machine learning techniques coupled with the growing accumulation of drug data have paved the way for the construction of computational models to predict drug clearance. However, concerns persist regarding the reliability of data collected from public sources, and a majority of current in silico quantitative structure-property relationship models tend to neglect the influence of molecular chirality. In this study, we meticulously examined human liver microsome (HLM) data from public databases and constructed two distinct data sets with varying HLM data quantity and quality. Two baseline models (RF and DNN) and three chirality-focused GNNs (DMPNN, TetraDMPNN, and ChIRo) were proposed, and their performance on HLM data was evaluated and compared with each other. The TetraDMPNN model, which leverages chirality from 2D structure, exhibited the best performance with a test R2 of 0.639 and a test root-mean-squared error of 0.429. The applicability domain of the model was also defined by using a molecular similarity-based method. Our research indicates that graph neural networks capable of capturing molecular chirality have significant potential for practical application and can deliver superior performance.

CTF-DDI: Constrained tensor factorization for drug–drug interactions prediction

Computational approaches for predicting drug–drug interactions (DDI) can significantly facilitate combination therapy and drug discovery. Existing similarity-based methods often overlook simple yet valuable structural information or ignore multiple relationships from biological entities (e.g., target proteins and enzymes). Meanwhile, matrix factorization-based methods can alleviate the inherent sparsity issues in DDI data. However, this line of work usually only considers the original association information of DDI pairs. To address these issues, we proposed a novel tensor factorization strategy with effective constraint terms (CTF-DDI) for potential DDI prediction. Specifically, we first obtained drug features by constructing specific similarity matrices based on drug structure and drug-related biological associations. Then, a novel constrained tensor factorization (CTF) module was designed to further reconstruct drug similarity by introducing Hessian and L2,1 regularization as constraints. Finally, we trained a deep neural network to extract nonlinear features for DDI prediction. Experimental results on two benchmark datasets demonstrated that the proposed CTF-DDI model outperforms classical tensor factorization and deep learning models. Furthermore, ablation and case studies validated the performance of CTF-DDI in DDI prediction. The source code of CTF-DDI is available at https://github.com/angelfacedac/CTF_DDI.

Unveiling the hidden dynamics of intermittent surface water: A remote sensing framework

Intermittent surface water frequently transitioning between water and land over months and years, plays a crucial and increasingly significant role in both social and ecological systems. However, their vital and dramatic dynamics have mainly remained invisible due to monitoring limitations. We present a new remote sensing framework to capture the long-term monthly dynamics of surface water bodies, applying it to Poyang Lake, the largest freshwater lake in China. This framework employed a random forest classifier on all available Landsat data to identify monthly surface water bodies. Additionally, we developed a Spatial and Temporal Neighborhood Similarity-based Gap Filling method to restore water bodies obscured by clouds and ensure spatial integrity. Furthermore, we introduced an index to quantify the intermittency of surface water bodies on a scale from 0 to 1, allowing for the classification of water bodies into three categories: perennial, wet intermittent, and dry intermittent. Employing this framework, we reconstructed the most complete monthly 30-m surface water dataset for cloudy regions to date, covering April 1986 to September 2023, demonstrating a strong correlation (Spearman's rank correlation coefficient of 0.909) with observed water levels. The results reveal a landscape dominantly composed of intermittent water bodies (91.2%), with a rapidly shrinking trend of perennial water bodies at 1303.58 ha per year. Notably, 162,685 ha (21.9%) of water bodies transitioned toward drier and more intermittent statuses. Dry intermittent water bodies exhibited the most pronounced land-water transitions, with the highest water-to-land (82.5%) and land-to-water (89.9%) proportions among the three categories. By uncovering the hidden dynamics of intermittent surface water, and highlighting its prevalence, expansion, and vulnerability, this framework paves the way for a better understanding of these critical water dynamics across the globe.

Slow feature-based feature fusion methodology for machinery similarity-based prognostics

Similarity-based prediction methods utilize degradation trend analysis based on degradation indicators (DIs). These methods are gaining prominence in industrial predictive maintenance because they effectively address prognostics for machines with unknown failure mechanisms. However, current studies often neglect the discrepancies in degradation trends when constructing DIs from multi-sensor data and lack automatic normalization of operating regimes during feature fusion. In this study, a feature fusion methodology based on a signal-to-noise ratio metric that leverages slow feature analysis (SFA) is proposed. This customized metric utilizes SFA to quantify degradation trend discrepancies of constructed DIs, while automatically filtering out the effects of multiple operating regimes during feature fusion. The effectiveness and superiority of the proposed method are demonstrated using publicly available aero-engine and rolling bearing datasets.

Similarity-based Graph Convolution Collaborative Recommendation Approach

With the rapid and iterative development of science and technology, a large amount of information exceeds the range that can be accepted, processed, or effectively utilized by an individual or a system, and recommendation algorithms can, to a certain extent, solve such problems, but traditional recommendation algorithms do not have a good solution to the problems related to data sparsity and recommendation accuracy. A similarity-based graph convolutional neural collaborative recommendation method (GCSCF) is proposed. The similarity algorithm based on the attributes of the item is used to find the item with the highest similarity and has not interacted with the current user, and this item is set to interact with the current user. The relevant interaction information of the user and the item is converted into relative feature vectors; the feature vectors are propagated using a graph convolutional neural network to aggregate the localized information, and the weight coefficients based on the item ratings are normalized to reduce the noise caused by the information aggregation. Comparative experiments are conducted on two public datasets, MovieLens-1M and Movielens-100K, with five baseline models on the set, and using Recall, Normalized Discounted Cumulative Gain (NDCG), and Precision as the evaluation metrics, and the results of the experiments show that the performance of the proposed social recommendation model better than other models.

Long-term (2013–2022) mapping of winter wheat in the North China Plain using Landsat data: classification with optimal zoning strategy

ABSTRACT Long-term mapping of winter wheat is vital for assessing food security and formulating agricultural policies. Landsat data are the only available source for long-term winter wheat mapping in the North China Plain due to the fragmented landscape in this area. Although various methods, such as index-based methods, curve similarity-based methods and machine learning-based methods, have been developed for winter wheat mapping based on remote sensing, the former two often require satellite data with high temporal resolution, which are unsuitable for Landsat data with sparse time-series. Machine learning is an effective method for crop classification using Landsat data. Yet, applying machine learning for winter wheat mapping in the North China Plain encounters two main issues: 1) the lack of adequate and accurate samples for classifier training; and 2) the difficulty of training a single classifier to accomplish the large-scale crop mapping due to the high spatial heterogeneity in this area. To address these two issues, we first designed a sample selection rule to build a large sample set based on several existing crop maps derived from recent Sentinel data, with specific consideration of the confusion error between winter wheat and winter rapeseed in the available crop maps. Then, we developed an optimal zoning method based on the quadtree region splitting algorithm with classification feature consistency criterion, which divided the study area into six subzones with uniform classification features. For each subzone, a specific random forest classifier was trained and used to generate annual winter wheat maps from 2013 to 2022 using Landsat 8 OLI data. Field sample validation confirmed the high accuracy of the produced maps, with an average overall accuracy of 91.1% and an average kappa coefficient of 0.810 across different years. The derived winter wheat area also has a good correlation (R2 = 0.949) with census area at the provincial level. The results underscore the reliability of the produced annual winter wheat maps. Additional experiments demonstrate that our proposed optimal zoning method outperforms other zoning methods, including Köppen climate zoning, wheat planting zoning and non-zoning methods, in enhancing wheat mapping accuracy. It indicates that the proposed zoning is capable of generating more reasonable subzones for large-scale crop mapping.

PhyloBARCODER: A Web Tool for Phylogenetic Classification of Eukaryote Metabarcodes Using Custom Reference Databases.

We developed phyloBARCODER (https://github.com/jun-inoue/phyloBARCODER), a new web tool that can identify short DNA sequences to the species level using metabarcoding. phyloBARCODER estimates phylogenetic trees based on the uploaded anonymous DNA sequences and reference sequences from databases. Without such phylogenetic contexts, alternative, similarity-based methods independently identify species names and anonymous sequences of the same group by pairwise comparisons between queries and database sequences, with the caveat that they must match exactly or very closely. By putting metabarcoding sequences into a phylogenetic context, phyloBARCODER accurately identifies (i) species or classification of query sequences and (ii) anonymous sequences associated with the same species or even with populations of query sequences, with clear and accurate explanations. Version 1 of phyloBARCODER stores a database comprising all eukaryotic mitochondrial gene sequences. Moreover, by uploading their own databases, phyloBARCODER users can conduct species identification specialized for sequences obtained from a local geographic region or those of nonmitochondrial genes, e.g. ITS or rbcL.

Towards Faithful Model Explanation in NLP: A Survey

Abstract End-to-end neural Natural Language Processing (NLP) models are notoriously difficult to understand. This has given rise to numerous efforts towards model explainability in recent years. One desideratum of model explanation is faithfulness, that is, an explanation should accurately represent the reasoning process behind the model’s prediction. In this survey, we review over 110 model explanation methods in NLP through the lens of faithfulness. We first discuss the definition and evaluation of faithfulness, as well as its significance for explainability. We then introduce recent advances in faithful explanation, grouping existing approaches into five categories: similarity-based methods, analysis of model-internal structures, backpropagation-based methods, counterfactual intervention, and self-explanatory models. For each category, we synthesize its representative studies, strengths, and weaknesses. Finally, we summarize their common virtues and remaining challenges, and reflect on future work directions towards faithful explainability in NLP.

Prediction of peptide hormones using an ensemble of machine learning and similarity-based methods.

Peptide hormones serve as genome-encoded signal transduction molecules that play essential roles in multicellular organisms, and their dysregulation can lead to various health problems. In this study, we propose a method for predicting hormonal peptides with high accuracy. The dataset used for training, testing, and evaluating our models consisted of 1174 hormonal and 1174 non-hormonal peptide sequences. Initially, we developed similarity-based methods utilizing BLAST and MERCI software. Although these similarity-based methods provided a high probability of correct prediction, they had limitations, such as no hits or prediction of limited sequences. To overcome these limitations, we further developed machine and deep learning-based models. Our logistic regression-based model achieved a maximum AUROC of 0.93 with an accuracy of 86% on an independent/validation dataset. To harness the power of similarity-based and machine learning-based models, we developed an ensemble method that achieved an AUROC of 0.96 with an accuracy of 89.79% and a Matthews correlation coefficient (MCC) of 0.8 on the validation set. To facilitate researchers in predicting and designing hormone peptides, we developed a web-based server called HOPPred. This server offers a unique feature that allows the identification of hormone-associated motifs within hormone peptides. The server can be accessed at: https://webs.iiitd.edu.in/raghava/hoppred/.

Local instance-based transfer learning for reinforcement learning

Similarity-based transfer learning for reinforcement learning has garnered attention for its potential to enhance target task learning. However, it faces significant challenges in efficiency and effectiveness, primarily stemming from issues such as sparse reward, long trajectory, and strict similarity. To solve these problems, this paper proposes a local instance-based transfer learning method for reinforcement learning. Instead of relying on sparse reward and long trajectory, this approach leverages the Q value of the local trajectory to evaluate similarity, thereby significantly enhancing transfer efficiency. Furthermore, by relaxing the strictness of the similarity, three transfer policies are proposed to facilitate positive transfer. Extensive experimental results demonstrate that the effectiveness and efficiency of the proposed method in comparison with traditional similarity-based transfer learning methods.

Phishing webpage detection based on global and local visual similarity

In recent years, phishing websites have constantly evolved, causing traditional URL or HTML-based detection methods less effective. This limitation motivated the development of visual similarity-based detection methods. These methods can identify phishing websites in a rapidly changing phishing landscape since the key factor in deceiving users is the high visual similarity between phishing pages and their corresponding legitimate ones. However, existing similarity-based methods often suffer from high false positive rates. To alleviate the problem, this paper proposed a new webpage similarity phishing detection framework. Using a deep learning model, the framework locates a website logo and integrates the logo area with the full webpage to detect phishing pages. Experimental evaluations on multiple datasets demonstrate the model’s ability to achieve a precision of 91% and a recall of 88% in detecting phishing websites. Comparative experiments against existing methods reveal that the proposed model enhances precision and recall by 17% and 21%, respectively, while reducing the false positive rate by 1%.

Similarity-based active learning methods

Active Learning has been a popular method to circumvent the labeling cost in machine learning methods. The majority of active learning approaches can be classified into two categories: representative-based and informative-based methods, with some hybrid methods that combine both. This work presents a naïve query strategy, namely Similarity-Based Active Learning (SBAL), which computes the sum of a row in the similarity matrix at each selection step, and a general optimization framework that can accommodate a broad range of active learning algorithms. The label complexity for different classification metrics is used as a primary criterion for comparing different algorithms. The proposed algorithm’s numerical performance is illustrated using simulated data scenarios and by applying it to the real-world COVID-19 image classification. The results demonstrate that, based on the classification metric, labeling cost and label complexity, SBAL outperforms other hybrid methods, such as Adaptive Active Learning (AAL) and Maximizing Variance for Active Learning (MVAL).

SSGAN: A Semantic Similarity-Based GAN for Small-Sample Image Augmentation

Image sample augmentation refers to strategies for increasing sample size by modifying current data or synthesizing new data based on existing data. This technique is of vital significance in enhancing the performance of downstream learning tasks in widespread small-sample scenarios. In recent years, GAN-based image augmentation methods have gained significant attention and research focus. They have achieved remarkable generation results on large-scale datasets. However, their performance tends to be unsatisfactory when applied to datasets with limited samples. Therefore, this paper proposes a semantic similarity-based small-sample image augmentation method named SSGAN. Firstly, a relatively shallow pyramid-structured GAN-based backbone network was designed, aiming to enhance the model’s feature extraction capabilities to adapt to small sample sizes. Secondly, a feature selection module based on high-dimensional semantics was designed to optimize the loss function, thereby improving the model’s learning capacity. Lastly, extensive comparative experiments and comprehensive ablation experiments were carried out on the “Flower” and “Animal” datasets. The results indicate that the proposed method outperforms other classical GANs methods in well-established evaluation metrics such as FID and IS, with improvements of 18.6 and 1.4, respectively. The dataset augmented by SSGAN significantly enhances the performance of the classifier, achieving a 2.2% accuracy improvement compared to the best-known method. Furthermore, SSGAN demonstrates excellent generalization and robustness.